SVGA-Net: Sparse Voxel-Graph Attention Network for 3D Object Detection from Point Clouds

Benchmark Model Rank Results
3d-object-detection-on-kitti-cars-easySVGA-Net–AP: 87.33%
3d-object-detection-on-kitti-cars-easy-valSVGA-Net–AP: 90.59
3d-object-detection-on-kitti-cars-hardSVGA-Net–AP: 74.63%
3d-object-detection-on-kitti-cars-hard-valSVGA-Net–AP: 79.15
3d-object-detection-on-kitti-cars-moderate-valSVGA-Net–AP: 80.23
3d-object-detection-on-kitti-cyclists-easySVGA-Net–AP: 79.22%
3d-object-detection-on-kitti-cyclists-hardSVGA-Net–AP: 57.64%
3d-object-detection-on-kitti-cyclists-moderateSVGA-Net–AP: 66.13%
3d-object-detection-on-kitti-pedestrians-easySVGA-Net–AP: 55.21%
3d-object-detection-on-kitti-pedestrians-hardSVGA-Net–AP: 44.56%
3d-object-detection-on-kitti-pedestrians-moderateSVGA-Net–AP: 47.71%